The effect of SNP calling criteria on the variance when calling genotypes. Comparison of the difference between our estimated Tajima’s D and the known Tajima’s D. These plots are based on a scenario with depth 2× and error rate 0.5% and show the difference of different p-values used for the LRT test. Left figure is the selection dataset and right figure is the neutral dataset. Notice that the 10-6 cutoff has quite the same variance in both plots. We observe that the 10-3 cutoff has less variance on the selection dataset, but more variance in the neutral dataset. We see the opposite correlation with regards to the 5 × 10-5 cutoff.